Compute Comparison
NVIDIAAmpere2020

RTX A6000

See rental prices ↓

Current median: $0.72/hr across 15 providers (observed), 0.00 (0.0%) vs yesterday. Typical range: $0.46–$1.67/hr (P10–P90). 30-day trend unlocks after 30 days of captures

TDP 300W · 77.4 TFLOPS · 48GB VRAM · 768 GB/s bandwidth

The RTX A6000 has a 300W TDP, 77.4 TFLOPS of FP16 compute, 48GB of GDDR6 VRAM, 768 GB/s memory bandwidth.

Ampere flagship professional GPU. 48GB GDDR6 with NVLink 3.0. Predecessor to RTX 6000 Ada. Widely available at lower cost.

VRAM
48GB
GDDR6
FP16
77.4
TFLOPS
Bandwidth
768.0
GB/s
TDP
300W
power
Best for:Large model inferenceMulti-GPU NVLink setupsProfessional rendering + AI

RTX A6000 currently rents for a median of $0.72/hr across 15 providers (observed on-demand prices). RTX A6000 has 48GB of memory. Compare its specs and listed cloud rental prices below. The workload must fit in VRAM with room for runtime memory; a listed price does not confirm stock.

RTX A6000 Overview

The RTX A6000 is an Ampere professional GPU with 48GB of ECC GDDR6, 77.4 TFLOPS of FP16/BF16 performance, and a 300W power envelope. As the predecessor to RTX 6000 Ada, it trades newer architecture performance for mature professional drivers, large memory, and broad availability.

Its 768 GB/s memory bandwidth is adequate for many workstation and inference workloads, while 48GB gives much more room than 24GB consumer cards for model weights, assets, and batch state. NVLink 3.0 allows paired professional cards to work together, but GDDR6 bandwidth remains a constraint compared with A100-class HBM platforms.

The RTX A6000 works well for professional rendering plus AI, larger-model inference, and cost-conscious multi-GPU workstation setups. It lacks FP8 support and does not match modern Ada or Hopper cards for pure inference throughput, so it is best chosen for capacity and professional features rather than speed alone.

Memory

VRAM48 GB
Memory TypeGDDR6
Bandwidth768 GB/s
NVLink BW112 GB/s

Compute Performance

FP3238.7 TFLOPS
FP1677.4 TFLOPS
BF1677.4 TFLOPS
INT8309 TOPS

Hardware Specifications

Chip

ArchitectureGA102
GenerationAmpere
Process NodeSamsung 8nm
Transistors28.3B
Die Size628 mm²
Release DateNovember 17, 2020
Launch MSRP$4,650

Processors

CUDA / Shader Cores10,752
Tensor Cores336
RT Cores84

Clocks

Base Clock1,455 MHz
Boost Clock1,860 MHz

Memory

VRAM48 GB
Memory TypeGDDR6
Memory Bus384-bit
Bandwidth768 GB/s
L2 Cache6 MB
NVLink BW112 GB/s

Power

TDP300 W
InterconnectNVLink 3.0 / PCIe 4.0

Relative Performance

FP16 Compute1%
VRAM Capacity17%
Mem Bandwidth5%

Relative to highest-spec GPU in database

Limitations

GDDR6 memory bandwidth far below HBM alternatives
No NVLink on A4000/A2000 — single-card VRAM ceiling
Professional but older Ampere architecture — no FP8 support

Live Cloud PricingOn-demand hourly rates

Live17 providers · -70% price spread
ProviderOn-DemandSpotRegionAvailability
$2.30/hr—EU/USunknown
$0.63/hr—EU-Westunknown
$0.79/hr—EUunknown
$0.72/hr—USunknown
$0.80/hr—ISunknown
$0.88/hr—EUunknown
$0.85/hr—EUunknown
$0.82/hr—SEunknown
$0.57/hr—Montreal, Canada, New York, NY, United States, Wichita, KS, United States, Kansas City, KS, United States, Des Moines, IA, United States, Las Vegas, NV, United Statesunknown
$2.20/hr—Variousunknown
$0.59/hr—USunknown
$0.45/hr—USunknown
$0.70/hr—APACunknown
$0.48/hr—US/Midwestunknown
$0.79/hr—EU-Northunknown
$0.35/hr—USunknown
$0.69/hr—EU-Northunknown
Cheapest / month
$1679
$2.30/hr × 730h
Median / month
$416
at median price
Spot savings
N/A
vs on-demand (best)

Job Cost Estimator

At median $0.72/hr across 15 providers (observed on-demand). Spot pricing typically 40–70% lower.

7B LLM fine-tune
8× GPU, 24 h · Full fine-tune, single node
$138
spot ~$62
70B LLM fine-tune
64× GPU, 72 h · Multi-node, 3-day run
$3.3k
spot ~$1.5k
Inference server
1× GPU, 730 h · 1 month continuous
$526
spot ~$237
Image gen batch
4× GPU, 8 h · Batch generation job
$23
spot ~$10

Estimates only. Actual cost depends on provider, region, availability, and utilization. Use the GPU cost calculator for custom inputs.

Compare RTX A6000 vs…

Use Case Guidance

Large model inference
Multi-GPU NVLink setups
Professional rendering + AI

LLM Model Size Guidance

Max model (FP16)~24Bparameters at FP16 precision
Max model (INT8)~48Bparameters at INT8 precision
Max model (INT4)~96Bparameters at INT4/GGUF

Estimates only. Actual capacity depends on context length, KV cache, and framework overhead.

Related Guides

LLM APIs Running on This GPU Class

Providers that serve frontier LLM inference on Ampere-class hardware.

Browse all 42 LLM models

RTX A6000 vs Alternatives — Spec Comparison

SpecRTX A6000 thisA40RTX 4070 TiRTX 3090
VRAM48GB GDDR6▲48GB GDDR6▲12GB GDDR6X24GB GDDR6X
Memory Bandwidth768 GB/s696 GB/s504 GB/s936 GB/s▲
FP16 TFLOPS77.474.880.2▲71
BF16 TFLOPS77.474.880.2▲71
FP8 TFLOPS————
INT8 TOPS309▲149.7160142
TDP300W300W285W350W▲
Process NodeSamsung 8nm▲Samsung 8nm▲TSMC 4NSamsung 8nm▲
ArchitectureGA102GA102AD104▲GA102
Release Year202020202023▲2020
Max model (FP16)~24B params▲~24B params▲~6B params~12B params
Max model (INT4)~96B params▲~96B params▲~24B params~48B params
▲ indicates best value in row · FP16/BF16 TFLOPS at full precision · Max model estimates at 2 bytes/param (FP16) and 0.5 bytes/param (INT4)Full side-by-side comparison

Compare RTX A6000

Related GPUs

RTX A6000 Frequently Asked Questions

What is the RTX A6000 TDP?

The RTX A6000 has a TDP (Thermal Design Power) of 300W.

How many TFLOPS does the RTX A6000 have?

The RTX A6000 delivers 77.4 TFLOPS of FP16 compute and 77.4 TFLOPS BF16.

How much VRAM does the RTX A6000 have?

The RTX A6000 has 48GB of GDDR6 VRAM.

What is the RTX A6000's memory bandwidth?

The RTX A6000 has a memory bandwidth of 768 GB/s.

RTX A6000 User Reviews

RTX A6000 User Reviews

Loading reviews...

Ready to rent?

Compare RTX A6000 prices across 130+ providers

Live on-demand & spot rates · monthly cost estimates · availability status

Compare rental prices